What we did
We chose three municipalities in the Vale do Sousa and mapped every active dental and aesthetic medicine clinic — 97 in total, cross-referencing the public register of healthcare establishments with location data.
Then we asked the questions a patient asks. Not technical marketing questions: the real ones.
What is the best dental clinic in [municipality]?
Where can I get dental implants in [municipality] with a good reputation?
Six questions, three platforms, eighteen measurements. All on the same day, in an anonymous session, with no history and no location sharing — so that the result would not be contaminated by personalisation.
We recorded who appears, in what order, and which sources each engine consulted to answer. It is this last part that explains everything else.
Five things we found
1. In each municipality, two or three clinics dominate. The rest don't exist.
There is no gradual distribution. There is a handful of names that appear consistently on all three platforms, and then a void.
In Paços de Ferreira, Clínica Médico Dentária Susana Mota was cited by all three platforms for the generic question, and Instituto Médico e Dentário Dra. Sara Martins by all three for the implants question. In Paredes, Clínica Dentária de Paredes and Clínica Médico Dentária Vale do Sousa. In Lousada, Clínica Sónia Costa and Coralis Clinics.
That is six clinics. The other 91 share isolated citations — or none.
2. The number of Google reviews determines nothing.
This was the most counter-intuitive result.
A clinic with three Google reviews was recommended by ChatGPT. Another with eight came first in Perplexity for implants. And a clinic with 37 reviews and a positive rating was recommended by none of the three — even though its Facebook page was read as a source.
Reputation among patients and visibility in AI engines are two different things. Having one doesn't give you the other.
3. Generic questions and specific questions have different winners.
Whoever wins “best dental clinic” is not whoever wins “where to get implants”. And the reason lies in the sources.
The clinic that leads the generic question has something no other clinic in the municipality has: an editorial article published about it on a content portal. Not a directory listing — a piece of writing.
The ones that lead the implants question have something else: a page on their own website whose title is the treatment plus the municipality. Literally “Dental Implants in [municipality]”.
Two different mechanisms, two different results. That is why testing only one question tells you nothing.
4. There is so little Portuguese content that AI turns to Brazil.
For the implants question in Paredes, three of the ten sources cited by Perplexity were websites of Brazilian clinics.
This is not a model error. It is the consequence of there not being enough Portuguese content on the subject. The engine needs to answer and uses what it finds.
For anyone who wants to occupy that space, it is the definition of opportunity.
5. Well-built content spills over into neighbouring municipalities.
The website of a clinic in Paços de Ferreira was cited as a source for a question about Lousada.
Its pages are structured so clearly — treatment, location, direct answer — that the engine considers it relevant beyond the area where it is based. This means visibility in AI is not limited to the municipality: it is built, and it spreads.
Why some appear and others don't
The answers these engines give don't come out of nowhere. They are built from a restricted set of sources, and we recorded which ones. The most cited, in this order:
- Doctoralia — the number one source in all three municipalities, without exception
- The clinics' own websites — but only those with dedicated pages per treatment
- saudebemestar.pt — individual listings per clinic
- Facebook and Instagram
- Health insurance networks — Medicare, AdvanceCare, Multicare, ADSE
- Websites of national chains — which dominate through domain authority
- Editorial content — rare, and therefore decisive
The pattern that emerges is simple: the clinics that are cited are present in five or more of these sources, with complete profiles. The invisible ones are in two or three, always with profiles created by third parties and never claimed.
It isn't luck. It's structure.
What this means for a clinic
A growing share of patients put these questions to an AI before searching on Google. And there is a fundamental difference between the two.
Google returns a list where you can compare, scroll down, explore. AI returns three to five names and nothing more.
Whoever isn't on that list isn't compared. They are ignored.
And there is an honest note that is part of this study: an independent clinic is not going to beat a national chain on the generic question in the short term. The chains have domain authority built up over years and a systematic presence in every source.
But it can win the questions about a municipality and a specific treatment — which is exactly where the patient who really intends to book is. That is the field where the contest is possible, and it is the one almost nobody is playing.
Methodology
So that the study can be repeated and verified.
| Population | 97 active dental and aesthetic medicine clinics in the municipalities of Paços de Ferreira, Paredes and Lousada |
|---|---|
| Population sources | Public register of healthcare establishments, cross-referenced with location data and sector directories |
| Questions | 6 — three on category and geography, three on a specific treatment |
| Platforms | ChatGPT, Google Gemini, Perplexity |
| Measurements | 18 |
| Date | 29 July 2026 |
| ChatGPT and Perplexity | Anonymous session, no history, no location sharing |
| Gemini | Signed in to an account used exclusively for measurements, with activity and personalisation turned off. The signed-out version does not return reliable results. |
| Record | Dated screenshot of every answer |
AI engine results vary over time and between users. This study is a snapshot of one day, with documented conditions — not a definitive measurement.
Frequently asked questions
What is GEO?
GEO stands for Generative Engine Optimisation: the work of making a brand citable in the answers of artificial intelligence engines such as ChatGPT, Gemini and Perplexity. It complements SEO — SEO works on positions in search results, GEO works on mentions inside generated answers.
How many dental clinics appear in ChatGPT's answers?
In the study we carried out in July 2026, with 97 clinics in the municipalities of Paços de Ferreira, Paredes and Lousada, only 6 were cited by all three platforms tested. Around 84% received no mention at all.
Does the number of Google reviews influence visibility in AI?
Not decisively. In our study, clinics with 3 and with 8 reviews were recommended, while a clinic with 37 reviews was not mentioned by any platform. The structure of the published information carries more weight than the volume of reviews.
Which sources does ChatGPT consult to recommend a dental clinic?
In our measurements, the most cited sources were Doctoralia, the clinics' own websites with dedicated pages per treatment, the saudebemestar.pt portal, social networks, health insurance networks and, occasionally, editorial content published about the clinic.
What can a clinic do to appear in AI answers?
The action with the most impact and the least effort is to claim and complete the profile in the directories these engines cite, starting with Doctoralia. After that, create a page per treatment on your own website with the treatment and the town in the title. Both are possible without outside help.
How was this study carried out?
We tested six questions on three AI platforms, for a total of 18 measurements, all on the same day and under documented conditions: an anonymous session in ChatGPT and Perplexity, and a dedicated account with no personalisation in Gemini. We took screenshots of every answer.